Genome-wide identification of directed gene networks using large-scale population genomics data

Abstract

Identification of causal drivers behind regulatory gene networks is crucial in understanding gene function. We developed a method for the large-scale inference of gene-gene interactions in observational population genomics data that are both directed (using local genetic instruments as causal anchors, akin to Mendelian Randomization) and specific (by controlling for linkage disequilibrium and pleiotropy). The analysis of genotype and whole-blood RNA-sequencing data from 3,072 individuals identified 49 genes as drivers of downstream transcriptional changes (P < 7 × 10−10), among which transcription factors were overrepresented (P = 3.3 × 10−7). Our analysis suggests new gene functions and targets including for SENP7 (zinc-finger genes involved in retroviral repression) and BCL2A1 (novel target genes possibly involved in auditory dysfunction). Our work highlights the utility of population genomics data in deriving directed gene expression networks. A resource of trans-effects for all 6,600 genes with a genetic instrument can be explored individually using a web-based browser.

Genome-wide identification of genes regulating DNA methylation using genetic anchors for causal inference

“…The majority (86%) of the target CpGs were located on the q-arm of chromosome 19. We found that for most of these CpGs (92%) their DNA methylation levels were associated with the expression levels of one or more nearby zinc fingers (table S13-14), consistent with previous gene network analyses (Luijk et al, 2018). Although SENP7 has no DNA-binding properties, previous research has shown that it exerts its effect through deSUMOylation of the chromatin repressor KAP1 (Garvin et al, 2013).…”

“…Second, closer inspection of proteins that do not have DNA-binding properties suggests that they may regulate DNA methylation through protein-protein interactions. Possible mechanisms include post-translational regulation (NFKBIE encodes for IκBε which retains NF-kB in the cytoplasm (Bonizzi and Karin, 2004)), post-translational modification (SENP7: deSUMOylates the repressor KAP1 (Luijk et al, 2018)) and recruitment of epigenetic proteins to specific target sites through association with a DNA-binding protein (NLRC5 associates with a protein complex in MHC-I region (Kobayashi and van den Elsen, 2012)). Third, a fraction of the identified genes overlap with genes in a database that focuses on the core epigenetic regulators (i.e.…”

“…In order to identify genes that influence DNA methylation, we employed an approach that consists of two parts. First, we identified predictive genetic variants for the expression of each gene in our data, which we aggregated into single predictive scores termed genetic instruments (GIs) (Luijk et al, 2018). Second, we used these GIs as causal anchors to establish directed effects of gene expression on genome-wide DNA methylation levels while ensuring that these associations were specific by accounting for linkage disequilibrium (LD) and pleiotropy among neighboring GIs (see figure 1 for an overview of the successive steps in the analysis).…”

“…These analyses resulted in directed associations between 2,223 genes and 5,284 CpGs (Bonferroni correction, P < 1.4 x 10 -11 ; Table S2). Although directed, the associations resulting from this analysis may not be specific for a single gene as linkage disequilibrium (LD) and/or pleiotropy may result in GIs that are predictive of multiple neighboring genes (Luijk et al, 2018). We therefore adjusted all significant GI-CpG pairs for all neighboring GIs (<1Mb) to account for correlation induced by LD/pleiotropy among neighboring genes in order to identify the specific gene in a region driving the association.…”

“…We recently developed a method to identify directed and specific gene-gene interactions in population omics data (Luijk et al, 2018). Instead of using measured gene expression, this method builds upon previous work in which genetic variants were utilized as causal anchors for gene expression (Gamazon et al, 2015;Gusev et al, 2016).…”